Modeling method and device for evaluation model

Abstract

At a serving end, modeling samples from a number of modeling scenarios are separately collected, where each modeling sample includes a scenario variable and several basic variables, and where the scenario variable indicates a modeling scenario that the modeling sample belongs to. A modeling sample set is generated by merging the modeling samples. An evaluation model is trained based on modeling samples in the modeling sample set to generate a trained evaluation model, where the trained evaluation model is universal, and where the trained evaluation model is configured to produce a score applicable to multiple service scenarios.

What is claimed is: 
     
         1 . A computer-implemented method, comprising:
 separately collecting, at a serving end, modeling samples from a plurality of modeling scenarios, wherein each modeling sample includes a scenario variable and a plurality of basic variables, and wherein the scenario variable indicates a modeling scenario that the modeling sample belongs to;   generating a modeling sample set by merging the modeling samples; and   training an evaluation model based on modeling samples in the modeling sample set to generate a trained evaluation model, wherein the trained evaluation model is universal, and wherein the trained evaluation model is configured to produce a score applicable to multiple service scenarios.   
     
     
         2 . The computer-implemented method of  claim 1 , wherein the evaluation model is an additive model, and wherein the evaluation model is built by adding a first model portion formed by basic variables and a second model portion formed by scenario variables. 
     
     
         3 . The computer-implemented method of  claim 1 , wherein generating the modeling sample set includes:
 separately defining, for each modeling scenario, a plurality of risk events;   classifying the modeling samples into good samples and bad samples by determining whether each collected modeling sample includes at least one of the risk events; and   summarizing the collected modeling samples to generate a modeling sample set.   
     
     
         4 . The computer-implemented method of  claim 1 , further comprising:
 defining a training sample weight for each modeling scenario based on a number of modeling samples in each modeling scenario.   
     
     
         5 . The computer-implemented method of  claim 1 , further comprising:
 collecting target data from a specific service scenario, wherein the target data includes a scenario variable and a plurality of basic variables;   inputting the target data to the trained evaluation model; and   outputting a score for the target data, wherein the score is universal if the target data scored in multiple service scenarios, and wherein the score is not universal if the target data scored in the specific service scenario the target data belongs to.   
     
     
         6 . The computer-implemented method of  claim 5 , wherein the target data scored in the service scenario that the target data belongs to, and wherein the score of the trained evaluation model is a sum of the corresponding scores of the basic variables of the target data in the evaluation model and a score of the scenario variable of the target data in the evaluation model. 
     
     
         7 . The computer-implemented method of  claim 5 , wherein the target data scored in multiple service scenarios, and wherein the output score of the trained evaluation model is a sum of corresponding scores of the plurality of basic variables of the target data in the evaluation model. 
     
     
         8 . A non-transitory, computer-readable medium storing one or more instructions executable by a computer system to perform operations comprising:
 separately collecting, at a serving end, modeling samples from a plurality of modeling scenarios, wherein each modeling sample includes a scenario variable and a plurality of basic variables, and wherein the scenario variable indicates a modeling scenario that the modeling sample belongs to;   generating a modeling sample set by merging the modeling samples; and   training an evaluation model based on modeling samples in the modeling sample set to generate a trained evaluation model, wherein the trained evaluation model is universal, and wherein the trained evaluation model is configured to produce a score applicable to multiple service scenarios.   
     
     
         9 . The non-transitory, computer-readable medium of  claim 8 , wherein the evaluation model is an additive model, and wherein the evaluation model is built by adding a first model portion formed by basic variables and a second model portion formed by scenario variables. 
     
     
         10 . The non-transitory, computer-readable medium of  claim 8 , wherein generating the modeling sample set includes:
 separately defining, for each scenario, a plurality of risk events;   classifying the collected modeling samples into good samples and bad samples by determining whether each collected modeling sample includes at least one of the risk event; and   summarizing the collected modeling samples to generate a modeling sample set.   
     
     
         11 . The non-transitory, computer-readable medium of  claim 8 , the operations further comprising:
 defining a training sample weight for each modeling scenario based on a number of modeling samples in each modeling scenario.   
     
     
         12 . The non-transitory, computer-readable medium of  claim 8 , the operations further comprising:
 collecting target data from a specific service scenario, wherein the target data includes a scenario variable and a plurality of basic variables;   inputting the target data to the trained evaluation model; and   outputting a score for the target data, wherein the score is universal if the target data scored in multiple service scenarios, and wherein the score is not universal if the target data scored in the specific service scenario the target data belongs to.   
     
     
         13 . The non-transitory, computer-readable medium of  claim 12 , wherein the target data scored in the service scenario that the target data belongs to, and wherein the output score of the trained evaluation model is a sum of the corresponding scores of the basic variables of the target data in the evaluation model and a score of the scenario variable of the target data in the evaluation model. 
     
     
         14 . The non-transitory, computer-readable medium of  claim 12 , wherein the target data scored in multiple service scenarios, and wherein the output score of the trained evaluation model is a sum of corresponding scores of the plurality of basic variables of the target data in the evaluation model. 
     
     
         15 . A computer-implemented system, comprising:
 one or more computers; and   one or more computer memory devices interoperably coupled with the one or more computers and having tangible, non-transitory, machine-readable media storing one or more instructions that, when executed by the one or more computers, perform one or more operations comprising:
 separately collecting, at a serving end, modeling samples from a plurality of modeling scenarios, wherein each modeling sample includes a scenario variable and a plurality of basic variables, and wherein the scenario variable indicates a modeling scenario that the modeling sample belongs to; 
 generating a modeling sample set by merging the modeling samples; and 
 training an evaluation model based on modeling samples in the modeling sample set to generate a trained evaluation model, wherein the trained evaluation model is universal, and wherein the trained evaluation model is configured to produce a score applicable to multiple service scenarios. 
   
     
     
         16 . The computer-implemented system of  claim 15 , wherein the evaluation model is an additive model, and wherein the evaluation model is built by adding a first model portion formed by basic variables and a second model portion formed by scenario variables. 
     
     
         17 . The computer-implemented system of  claim 15 , wherein generating the modeling sample set includes:
 separately defining, for each scenario, a plurality of risk events;   classifying the collected modeling samples into good samples and bad samples by determining whether each collected modeling sample includes at least one of the risk event; and   summarizing the collected modeling samples to generate a modeling sample set.   
     
     
         18 . The computer-implemented system of  claim 15 , further comprising:
 defining a training sample weight for each modeling scenario based on a number of modeling samples in each modeling scenario.   
     
     
         19 . The computer-implemented system of  claim 15 , the operations further comprising:
 collecting target data from a specific service scenario, wherein the target data includes a scenario variable and a plurality of basic variables;   inputting the target data to the trained evaluation model; and   outputting a score for the target data, wherein the score is universal if the target data scored in multiple service scenarios, and wherein the score is not universal if the target data scored in the specific service scenario the target data belongs to.   
     
     
         20 . The computer-implemented system of  claim 19 , wherein the target data scored in multiple service scenarios, and wherein the output score of the trained evaluation model is a sum of corresponding scores of the plurality of basic variables of the target data in the evaluation model.